FUN3D v12.

4 Training
Session 3:
Gridding Considerations,
Solution Basics, and Visualization

Eric Nielsen

FUN3D Training Workshop

http://fun3d.larc.nasa.gov 1
March 24-25, 2014
 Learning Goals
What we will cover
• Basic gridding requirements and formats
• Nondimensionalizations and axis conventions
• Basic environment for running FUN3D
• FUN3D user inputs
• Running FUN3D for typical steady-state RANS cases
– Compressible transonic turbulent flow over a wing-body using a
tetrahedral VGRID mesh
– Turbulent flow over a NACA 0012 airfoil section
• Things to help diagnose problems
• Visualization overview

What we will not cover

• Other speed regimes
• Unsteady flows

FUN3D Training Workshop

http://fun3d.larc.nasa.gov 2
March 24-25, 2014
 Gridding Considerations
• FUN3D is a node-based discretization
– To get similar resolution when comparing with a cell-centered code, you must
use a finer grid
• E.g., on a tetrahedral grid, the grid for FUN3D must be ~2 times finer on the surface,
and ~6 times finer in the volume mesh to be fair
– This is critical when comparing with cell-centered solvers
– Hanging nodes are not currently supported
• FUN3D integrates all of the way to the wall for turbulent flows
– Wall function grids are not adequate
– Goal is to place first grid point at y+=1
• Base ∆y on a flat plate estimate using your Reynolds number; can examine result in
solver output and tweak as necessary
• Users employ all of the common grid generators – VGRID,
AFLR2/AFLR3/SolidMesh, ICEM, Pointwise, etc.
• FUN3D also supports point-matched, multiblock structured grids through
Plot3D file input
– Subject to certain grid topologies:
• Singularities treated – i.e., hexes with collapsed faces converted to prisms
• But hexes with 180° internal angles cause FUN3D discretization to break down (LSQ)
• FUN3D can convert tetrahedral VGRID meshes to mixed elements
• FUN3D can convert any mixed element grid into tetrahedra using command
line option ‘--make_tets’

FUN3D Training Workshop

http://fun3d.larc.nasa.gov 3
March 24-25, 2014
 Supported Grid Formats
Supports mixed Direct load or
Grid Format Formatted Unformatted File extension(s)
elements converter

FAST X X Direct .fgrid, .mapbc

VGRID X Direct .cogsg, .bc, .mapbc

(single or multisegment)

X .ugrid/.(l)r8.ugrid/.(l)b8.ugrid,
AFLR3 X X Direct
Also Binary .mapbc

FUN2D X Direct .faces

Fieldview v2.4, Direct .fvgrid_fmt, .fvgrid_unf,

X X X (Some details of format
v2.5, v3.0 not supported)
.mapbc

Felisa X Direct .gri, .fro, .bco

Point-matched, Hexes,
X X Converter .p3d, .nmf
multiblock Plot3D degenerates

CGNS Binary X Converter .cgns

The development team can work with you to handle other formats as needed

FUN3D Training Workshop

http://fun3d.larc.nasa.gov 4
March 24-25, 2014
 Boundary Condition Input File
• Where required, the FUN3D .mapbc file takes the form:

Number of boundary patches

Boundary patch index BC index Family name

• The BC index may be either a 4-digit FUN3D-style index or a GridTool-style

index
• The family name is optional, but must be present if the user requests patch
lumping by family

3
1 4000 Wing
2 5000 Farfield
3 6662 Symmetry plane

• Exception: The .mapbc format for VGRID meshes follows the

GridTool/VGRID format

FUN3D Training Workshop

http://fun3d.larc.nasa.gov 5
March 24-25, 2014
 Nondimensionalization
• Notation: * indicates a dimensional variable, otherwise dimensionless; the
reference flow state is usually free stream (“ ∞“), but need not be
• Define reference values:
– L*ref = reference length of the physical problem (e.g. chord in ft)
– Lref = corresponding length in your grid (dimensionless)
*
– ρ ref = reference density (e.g. slug/ft3)
*
– µ ref = reference molecular viscosity (e.g. slug/ft-s)
*
– T ref = reference temperature (e.g. oR, compressible only)
*
– a ref = reference sound speed (e.g. ft/s, compressible only)
*
– U ref = reference velocity (e.g. ft/s)
• Space and time are made dimensionless in FUN3D by:
 * *
– x = x /(Lref /Lref ) t = t * a*ref /( L*ref / Lref ) t = t *U ref
*
/( L*ref / Lref )
(compressible) (incompressible)
FUN3D Training Workshop
http://fun3d.larc.nasa.gov 6
March 24-25, 2014
 Nondimensionalization (cont)
• For the compressible flow equations the dimensionless variables are:
 * * *
u = u /aref = u ref /a*ref = M ref

– so u ref
* * *2 *
– P = P /( ρ ref aref ) so Pref = Pref /( ρ *ref a*2
ref ) = 1/ γ
* *
– a = a / aref so aref =1
– T = T * / Tref
*
so Tref =1
* * *2
– e = e /( ρ ref aref ) so eref = e*ref /( ρ *ref a*2
ref ) = 1/( γ ( γ −1)) + M 2
ref /2
* *
– ρ = ρ / ρ ref so ρ ref =1
– From the equation of state and the definition of sound speed:
T = γ P / ρ = a2
• The input Reynolds number in FUN3D is related to the Reynolds number
of the physical problem by
reynolds_number = Re ref /Lref where Re ref = ρ *ref U ref
*
L*ref / µ*ref
i.e. reynolds_number is a Reynolds number per unit grid length
FUN3D Training Workshop
http://fun3d.larc.nasa.gov 7
March 24-25, 2014
 Setting the Reynolds Number Input

• Frequent cause of confusion, even for developers

• Need to know what characteristic length your Reynolds number is
based on – mean aerodynamic chord, diameter, etc.
• Your input Reynolds number is based on the corresponding length
of that “feature” in your computational grid
• Example: You want to simulate a Reynolds number of 2.5 million
based on the MAC:
– If the length of the MAC in your grid is 1.0 grid units, you would input
Re=2500000 into FUN3D
– If the length of the MAC in your grid is 141.2 grid units (perhaps these
physically correspond to millimeters), you would input 2500000/141.2,
or Re=17705.4 into FUN3D

FUN3D Training Workshop

http://fun3d.larc.nasa.gov 8
March 24-25, 2014
 FUN3D Axis Convention

• FUN3D coordinate system differs from the standard wind coordinate system
by a 180º rotation about the y-axis
• Positive x-axis is toward the “back” of the vehicle (downstream)
• Positive y-axis is out the “right wing”
• Positive z-axis is “upward”
• The freestream angle of attack and yaw angle are defined as shown

FUN3D Training Workshop

http://fun3d.larc.nasa.gov 9
March 24-25, 2014
 Runtime Environment
• “Unlimit” your shell (also good idea to put this in any queue scripts):
$ ulimit unlimited # for bash
$ unlimit # for c shell
• If unformatted or binary, what “endianness” does your grid file have?
– E.g., VGRID files are always big endian, regardless of platform
– If your compiler supports it, FUN3D will attempt to open files using an
open(convert=…) syntax
– Most compilers support some means of conversion
• Either an environment variable or compile-time option, depending on what
compiler you’re using
• E.g., Intel compiler can be controlled with an environment variable
F_UFMTENDIAN = big
• Memory required by solver: rough rule of thumb is 3-3.5 GB per million
points (not cells!)
– Conversely, 200k-300k points per 1 GB of memory
• Users generally partition into smaller domains than this, but be aware of these
numbers
– This memory estimate will be higher if visualization options are used, etc
FUN3D Training Workshop
http://fun3d.larc.nasa.gov 10
March 24-25, 2014
 User Inputs for FUN3D
Input deck fun3d.nml
• The user is required to supply an input deck for FUN3D named fun3d.nml
(fixed name)
• This filename contains a collection of Fortran namelists that control FUN3D
execution – all namelist variables have default values as documented
• But user will need to set at least some high-level variables, such as the project
name

Command Line Options (CLOs)

• CLOs always take the form --command_line_option after the executable
name
– Some CLOs may require trailing auxiliary data such as integers and/or reals
• User may specify as many CLOs as desired
• CLOs always trump fun3d.nml inputs
• CLOs available for a given code in the FUN3D suite may be viewed by using
--help after the executable name
• Most CLOs are for developer use; namelist options are preferred where
available
FUN3D Training Workshop
http://fun3d.larc.nasa.gov 11
March 24-25, 2014
 Transonic Turbulent Flow on a
Tetrahedral Wing-Body Mesh

• For this case, we will assume that someone has provided a set of
VGRID files containing the mesh
– f6fx2b.cogsg, f6fx2b.bc, and f6fx2b.mapbc
• It is always a good idea to examine the .mapbc file first to check the
boundary conditions and any family names
– Note that specific boundary conditions will be covered in a separate session

FUN3D Training Workshop

http://fun3d.larc.nasa.gov 12
March 24-25, 2014
 Transonic Turbulent Flow on a
Tetrahedral Wing-Body Mesh
#Thu Mar 11 13:42:40 2010
#bc.map
Patch # BC Family #surf surfIDs Family
#---------------------------------------------------------------------
1 3 3 0 0 Box
• For this case, the VGRID/GridTool-style 2
3
3
3
3
3
0
0
0
0
Box
Box
4 3 3 0 0 Box
.mapbc file is as shown 5
6
3
4
3
4
0
1
0
15
Box
Wing
7 4 4 1 15 Wing
• Surface grid consists of 51 patches 8
9
4
4
4
4
1
1
17
17
Wing
Wing
10 4 4 1 15 Wing
• Note that VGRID/GridTool-style BC’s are 11
12
4
4
4
4
1
1
13
21
Fuselage
Fuselage
13 4 4 1 11 Fuselage
specified 14
15
4
4
4
4
1
1
11
12
Fuselage
Fuselage
16 4 4 1 12 Fuselage
• Family names are also as shown 17
18
4
4
4
4
1
1
15
15
Wing
Wing
19 4 4 1 15 Wing
(required in this format) 20
21
4
4
4
4
1
1
15
17
Wing
Wing
22 4 4 1 17 Wing
• FUN3D does not use the other columns 23
24
4
4
4
4
1
1
16
15
Wing
Wing
25 4 4 1 17 Wing
of data 26
27
4
4
4
4
1
1
8
16
Fuselage
Wing
28 4 4 1 16 Wing
• If you cannot easily visualize your mesh 29
30
4
4
4
4
1
1
16
16
Wing
Wing
31 4 4 1 18 Wing
to set appropriate boundary conditions, 32
33
4
4
4
4
1
1
18
17
Wing
Wing
34 4 4 1 18 Wing
one easy approach is to set them all to 35
36
4
4
4
4
1
1
18
1
Wing
Wing
37 4 4 1 18 Wing
inflow/outflow, then run a single time step 38
39
4
4
4
4
1
1
18
18
Wing
Wing
40 4 4 1 22 Fuselage
of FUN3D with boundary visualization 41
42
1
4
1
4
0
1
0
10
Symmetry
Fuselage
43 4 4 1 9 Fuselage
activated – then set patch BC’s as 44
45
4
4
4
4
1
1
14
23
Fuselage
Fuselage
46 4 4 1 19 Wing
needed for actual simulation 47
48
4
4
4
4
1
1
20
27
Wing
Fairing
49 4 4 1 29 Fairing
50 4 4 1 28 Fairing
51 4 4 1 30 Fairing

FUN3D Training Workshop

http://fun3d.larc.nasa.gov 13
March 24-25, 2014
 Transonic Turbulent Flow on a
Tetrahedral Wing-Body Mesh

• Now we will look at the minimum set of user inputs needed in

fun3d.nml to run this case

FUN3D Training Workshop

http://fun3d.larc.nasa.gov 14
March 24-25, 2014
 Transonic Turbulent Flow on a
Tetrahedral Wing-Body Mesh
&project
project_rootname = 'f6fx2b’ Project name
/
&raw_grid
grid_format = 'vgrid’ Read a set of VGRID files
/
&reference_physical_properties
mach_number = 0.75 Sets freestream Mach number
reynolds_number = 17705.40 Sets Reynolds number
angle_of_attack = 1.0 Sets freestream angle of attack
temperature = 580.0 Sets freestream temperature
temperature_units = "Rankine” Uses Rankine temperature units for input
/
&code_run_control
restart_read = 'off’ Perform a cold start
steps = 1000 Perform 1000 time steps
/
&force_moment_integ_properties
area_reference = 72700.0 Sets reference area
x_moment_length = 141.2 Sets length for normalizing y-moments
All in
y_moment_length = 585.6 Sets length for normalizing x-, z-moments
x_moment_center = 157.9 Sets x-moment center grid units
z_moment_center = -33.92 Sets z-moment center
/
&nonlinear_solver_parameters
schedule_cfl = 10.0 200.0 CFL for meanflow is ramped from 10.0 to 200.0
schedule_cflturb = 1.0 30.0 CFL for turbulence is ramped from 1.0 to 30.0
/

FUN3D Training Workshop

http://fun3d.larc.nasa.gov 15
March 24-25, 2014
 Transonic Turbulent Flow on a
Tetrahedral Wing-Body Mesh
• We now have the boundary conditions and input deck set up to run
FUN3D
• To execute FUN3D, we use the following basic command line syntax:
mpirun ./nodet_mpi
– Note your environment may require slightly different syntax:
• mpirun vs mpiexec vs aprun vs …
• May need to specify various MPI runtime options:
• -np #
• -machinefile filename
• -nolocal
• Others

FUN3D Training Workshop

http://fun3d.larc.nasa.gov 16
March 24-25, 2014
 Transonic Turbulent Flow on a
Tetrahedral Wing-Body Mesh
• Using 1 Xeon Ivy Bridge node (20 cores), this case runs in 2-3 minutes
• The top of the screen output will include an echo of your fun3d.nml, as well
as some preprocessing information:
FUN3D 12.4-69612M Flow started 02/22/2014 at 18:58:46 with 20 processes FUN3D version, start time, job size
[Echo of fun3d.nml]
The default "unformatted" data format is being
used for the grid format "vgrid". VGRID input is being used
rotor.input not found

moving_body.input not found

... nsegments,ntet,nnodesg 1 2994053 513095 Grid contains 2,994,053 tets and 513,095 points
cell statistics: type, min volume, max volume, max face angle Min/max cell volumes, max internal face angles
cell statistics: tet, 0.41152313E-06, 0.66593449E+11, 179.973678915
cell statistics: all, 0.41152313E-06, 0.66593449E+11, 179.973678915

... PM (64,skip_do_min) : 0 F
... Calling ParMetis (ParMETIS_V3_PartKway) .... 0 F
... edgeCut 104885 # of edges cut by partitioning (measure of communication)
... Time for ParMetis: .2 s
... Constructing partition node sets for level-0... 2994053 T
... Edge Partitioning ....
... Boundary partitioning....
... Reordering for cache efficiency....
... Write global grid information to f6fx2b.grid_info
... Time after preprocess TIME/Mem(MB): 1.60 180.52 180.52 1.6 secs required to preprocess the mesh
NOTE: kappa_umuscl set by grid: .00
Grid read complete
y-symmetry metrics modified/examined: 23418/23418
Distance_function unique ordering T 20000000
construct partial boundary...nloop= 1
find closer surface edge...
find closer surface face...

Wall spacing: 0.766E-03 min, 0.121E-02 max, 0.115E-02 avg Min/max/avg wall spacing statistics

FUN3D Training Workshop

http://fun3d.larc.nasa.gov 17
March 24-25, 2014
 Transonic Turbulent Flow on a
Tetrahedral Wing-Body Mesh
• At this point, time stepping commences
• For each time step:
– The L2-norm of the density|turbulence equation is red|blue; max and location are also included
– Lift and drag are reported in green
• “Done.” indicates execution is complete
Iter density_RMS density_MAX X-location Y-location Z-location
turb_RMS turb_MAX X-location Y-location Z-location
1 0.567457404772944E+00 0.28035E+02 0.16377E+03 -0.16562E+03 0.20117E+02
0.764159584901413E+04 0.13249E+07 0.79654E+04 -0.88280E+04 0.25675E+02
Lift 0.103226565173772E+00 Drag 0.646513396068887E+00
2 0.300679598726331E+00 0.12718E+02 0.29226E+03 -0.72487E+02 -0.12411E+02
0.753354470463467E+04 0.12868E+07 0.79654E+04 -0.88280E+04 0.25675E+02
Lift 0.146829230859457E+00 Drag 0.721243167013704E+00
.
.
.
999 0.383370843514542E-05 0.13909E-03 0.35380E+03 -0.58429E+02 -0.16200E+02
0.318320572426105E-02 0.19891E+00 0.36848E+03 -0.68458E+02 0.31074E+01
Lift 0.556387990643583E+00 Drag 0.388233647462313E-01
1000 0.382497896407724E-05 0.13871E-03 0.35380E+03 -0.58429E+02 -0.16200E+02
0.317436044959994E-02 0.19835E+00 0.36848E+03 -0.68458E+02 0.31074E+01
Lift 0.556387923023456E+00 Drag 0.388233658091165E-01

Writing f6fx2b.flow (version 11.8) lmpi_io 2

inserting current history iterations 1000
Time for write: .1 s
Done.

FUN3D Training Workshop

http://fun3d.larc.nasa.gov 18
March 24-25, 2014
 Transonic Turbulent Flow on a
Tetrahedral Wing-Body Mesh
• FUN3D provides a couple of text files with basic statistics and summary data:
– f6fx2b.grid_info File containing basic mesh statistics and partitioning info
– f6fx2b.forces File containing force breakdowns by boundary and totals
• FUN3D also produces:
f6fx2b_hist.dat Tecplot file with residual, force convergence histories
f6fx2b.flow Solver restart information

• For this particular case, the mean

flow and turbulence residuals are
reduced by ~5 and ~7 orders of
magnitude respectively over 1000
time steps
• Lift and drag come in after a few
hundred time steps

FUN3D Training Workshop

http://fun3d.larc.nasa.gov 19
March 24-25, 2014
 NACA 0012 Airfoil

• For this case, we have been given a set of

binary, big endian AFLR3 files
– 0012.b8.ugrid, 0012.mapbc 0012.mapbc
– For computations in 2D mode 4
• Grid must be one-element wide in the y-direction 1 4000
(except when using FUN2D format) 2 5000
• Grid must contain only prisms and/or hexes 3 6662
• First check the .mapbc file 4 6662
– The y-planes must be separate boundary patches
and should be given BC 6662
FUN3D Training Workshop
http://fun3d.larc.nasa.gov 20
March 24-25, 2014
 NACA 0012 Airfoil
&project
project_rootname = '0012’
/
• fun3d.nml is shown here &raw_grid
grid_format = 'aflr3’ Read an AFLR3 grid
• FUN2D grid format will data_format = 'stream’ Grid is in binary format
twod_mode = .true. Execute in 2D mode
automatically be executed /
in 2D mode; all others &reference_physical_properties
mach_number = 0.80
must be explicitly put in reynolds_number = 1.e6
2D mode angle_of_attack = 1.25
temperature = 580.0
temperature_units = "Rankine"
/
&code_run_control
restart_read = 'off'
steps = 5000
/
&force_moment_integ_properties
area_reference = 0.1
x_moment_center = 0.25
/
&nonlinear_solver_parameters
schedule_cfl = 10.0 200.0
schedule_cflturb = 1.0 10.0
/

FUN3D Training Workshop

http://fun3d.larc.nasa.gov 21
March 24-25, 2014
 NACA 0012 Airfoil
FUN3D 12.4-69634 Flow started 02/23/2014 at 10:50:52 with 20 processes
[Echo of fun3d.nml]
Preparing to read binary AFLR3 grid: 0012.b8.ugrid Binary AFLR3 grid being read

nnodes 116862 Grid contains 116,862 points

ntface,nqface 204510 14607 Grid contains 204,510 tris, 14,607 quads
ntet,npyr,nprz,nhex 0 0 102255 7047 Grid contains 102,255 prisms, 7,047 hexes

cell statistics: type, min volume, max volume, max face angle Cell stats now broken out by cell type
cell statistics: prz, 0.16960303E-06, 0.52577508E-01, 164.861624007
cell statistics: hex, 0.83173480E-09, 0.12843645E-04, 123.906431556
cell statistics: all, 0.83173480E-09, 0.52577508E-01, 164.861624007

... PM (64,skip_do_min) : 0 F
... Calling ParMetis (ParMETIS_V3_PartKway) .... 0 F
... edgeCut 10528
... Time for ParMetis: .0 s
... checking for spanwise edge cuts.
... Constructing partition node sets for level-0... 109302 T
... Edge Partitioning ....
... Boundary partitioning....
... Reordering for cache efficiency....
... ordering edges for 2D.
... Write global grid information to 0012.grid_info
... Time after preprocess TIME/Mem(MB): 0.29 61.23 61.23
NOTE: kappa_umuscl set by grid: .00
Grid read complete

Using 2D Mode (Node-Centered) Solver running in 2D mode

Distance_function unique ordering T 20000000

construct partial boundary...nloop= 1
find closer surface edge...
find closer surface face...

Wall spacing: 0.100E-03 min, 0.100E-03 max, 0.100E-03 avg

FUN3D Training Workshop

http://fun3d.larc.nasa.gov 22
March 24-25, 2014
 NACA 0012 Airfoil

FUN3D Training Workshop

http://fun3d.larc.nasa.gov 23
March 24-25, 2014
 List of Key Input/Output Files
• Input
– Grid files (prefixed with project name, suffixes depend on grid format)
– fun3d.nml
• Output
– project.grid_info
– project.forces
– project_hist.dat
– project.flow

FUN3D Training Workshop

http://fun3d.larc.nasa.gov 24
March 24-25, 2014
 What Could Possibly Go Wrong?
Problem
• Common complaint from VGRID meshes during initial preprocessing
phase at front end of solver:
Checking volume-boundary connectivity...

stopping...unable to find common element for face 1 of

boundary 3
boundary nd array 46 17368 334315

node,locvc 46******************************
node,locvc_type 46 tet tet tet tet tet

node,locvc 17368************************************
node,locvc_type 17368 tet tet tet tet tet tet

• This is due to a very old VGRID bug that causes an incompatibility

between the .cogsg and .bc files
– Compile and run utils/repair_vgrid_mesh.f90 to generate a valid
.bc file to replace your original one

FUN3D Training Workshop

http://fun3d.larc.nasa.gov 25
March 24-25, 2014
 What Could Possibly Go Wrong?
Problem
• Common complaint from unformatted/binary meshes during initial
preprocessing phase at front end of solver:
Read/Distribute Grid.
forrtl: severe (67): input statement requires too much data, unit 16100,
file /misc/work14/user/FUN3D/project.cogsg

• Check the endianness of the grid and your environment/executables

Problem
• Unexpected termination, especially during preprocessing or first time
step
– Are your shell limits set?
– Do you have enough local memory for what you are trying to run?

FUN3D Training Workshop

http://fun3d.larc.nasa.gov 26
March 24-25, 2014
 What Could Possibly Go Wrong?
Problem
• Solver diverges or does not converge
– Problem-dependent, very tough to give general advice here
– Sometimes require first-order iterations (primarily for high speeds)
– Sometimes require smaller CFL numbers
– Sometimes require alternate flowfield initialization (not freestream) in
some subregion of the domain: e.g., chamber of an internal jet
– Perhaps your problem is simply unsteady

Problem
• Solver suddenly dies during otherwise seemingly healthy run
– Sometimes useful to visualize solution just before failure
– Is it a viscous case on a VGRID mesh? Try turning on
large_angle_fix in &special_parameters namelist (viscous flux
discretization degenerates in sliver cells common to VGRID meshes)
– Is it a turbulent flow on a mesh generated using AFLR3? Look for
“eroded” boundary layer grids near geometric singularities – AFLR3
sometimes has trouble adding viscous layers near complex corners, etc
FUN3D Training Workshop
http://fun3d.larc.nasa.gov 27
March 24-25, 2014
 What Could Possibly Go Wrong?

In General…
• Do not hesitate to send questions to [email protected] ;
we are happy to try to diagnose problems
– Please send as much information about the problem/inputs/environment
that you can, as well as all screen output, any error output, and
config.log
– In extreme cases, we may request your grid and attempt to run a case for
you to track down the problem
– If you cannot send us a case due to restrictions, size, etc, a
generic/smaller representative case that behaves similarly can be useful
– Check the manual for guidance
• Ask the FUN3D user community, [email protected]

FUN3D Training Workshop

http://fun3d.larc.nasa.gov 28
March 24-25, 2014
 Visualization Learning Goals
• What this will teach you
– Run-time flow visualization output
• Output on boundary surfaces
• Output on user-specified “sampling” surfaces within the volume
• Output of full volume data
• Output generated by “slicing” boundary data - “sectional” output
• What you will not learn
– The plethora of output options available for visualization
– Tecplot usage
• What should you already know
– Basic flow solver operation and control

FUN3D Training Workshop

http://fun3d.larc.nasa.gov 29
March 24-25, 2014
 Background
• Datasets are getting simply too large to post-process in a traditional
manner
• FUN3D allows visualization data to be generated as the solver is
running
• User specified frequency and output type
• User specified output variables from a fairly extensive list
• Majority of output options are Tecplot-based
– Volume output may also be generated in Fieldview, CGNS formats
• Note FUN3D also supports true in-situ visualization at scale using the
DoE VisIt package; however, this is not covered here
– Intelligent Light is currently integrating VisIt’s in-situ capabilities
with Fieldview

FUN3D Training Workshop

http://fun3d.larc.nasa.gov 30
March 24-25, 2014
Selected Visualization Output Examples
Sliced Boundary Output
-4

-3.5

-3

-2.5 y/b = 0.08

y/b = 0.38
y/b = 0.68
-2 y/b = 0.99

Cp
-1.5

-1

-0.5

0.5

1
-0.5 -0.25 0 0.25 0.5 0.75 1
x

Iso-surfaces

Schlieren,
boundary output

FUN3D Training Workshop

http://fun3d.larc.nasa.gov 31
March 24-25, 2014
 Visualization Overview
• All of the visualization outputs require similar namelist-specified
“frequency” N to activate:
– In all cases, N = 0, 1, 2, 3, …
• N = 0 generates no output
• N < 0 generates output only at the end of the run - typically used for
steady-state cases. The actual value of N is ignored
• N > 0 generates output every Nth time step - typically used to
generate animation for unsteady flows; can also be used to observe
how a steady flow converges

FUN3D Training Workshop

http://fun3d.larc.nasa.gov 32
March 24-25, 2014
 Visualization Overview
• Customizable output variables (except sliced boundary data):
– Most variables are the same between the boundary surface, sampling
and volume output options; boundary surface has a few extra
– See manual for lists of all available variables
– Default variables always include x, y, z, and the “primitive” flow
variables u, v, w, and p (plus density if compressible)
– Several “shortcut” variables: e.g. primitive_variables = rho, u, v, w, p
– Must explicitly turn off the default variables if you don’t want them (e.g.
primitive_variables = .false.)
– Variable selection for each co-processing option done with a different
namelist to allow “mix and match”

FUN3D Training Workshop

http://fun3d.larc.nasa.gov 33
March 24-25, 2014
 Visualization Overview
• For boundary surface output, default is all solid boundaries in 3D and one
y=const. plane in 2D; alternate output boundaries are selected with (e.g.)
&boundary_output_variables
number_of_boundaries = 3
boundary_list = ‘3,5,9’ ! blanks OK as delimiter too: ‘3 5 9’
! dashes OK as delimiter too: ‘3-9’
/

• If you already have a converged solution and don’t want to advance the
solution any further, can do a “pass through” run:
– set steps = 0 in &code_run_control
– You must have a restart file ([project].flow)
– Run the solver with the appropriate namelist input to get desired
output
– [project].flow will remain unaltered after completion

FUN3D Training Workshop

http://fun3d.larc.nasa.gov 34
March 24-25, 2014
 Visualization Overview
• Sampling output requires additional data to describe the desired sampling
surface(s)
– Specified in namelist &sampling_parameters
– Surfaces may be planes, quadrilaterals or circles of arbitrary
orientation, or may be spheres or boxes
– Isosurfaces and schlierens also available
– Points may also be sampled
– See manual for complete info
• Sliced boundary surface output requires additional data to describe the
desired slice section(s)
– Specified in namelist &slice_data
– Always / only outputs x, y, z, Cp, Cfx, Cfy, Cfz
– User specifies which (solid) boundaries to slice, and where
– See manual for complete info

FUN3D Training Workshop

http://fun3d.larc.nasa.gov 35
March 24-25, 2014
 Visualization Overview
• Output files will be ASCII unless you have built FUN3D against the
Tecplot library (exception: sliced boundary data is always ASCII)
– ASCII files have .dat extension
– Binary files have .plt extension - smaller files; load into Tecplot faster
– Boundary output file naming convention (T = time step counter):
• [project]_tec_boundary_timestepT.dat if N > 0
• [project]_tec_boundary.dat if N < 0
– Volume output file naming convention (note: 1 file per processor P)
• [project]_partP_tec_volume_timestepT.dat if N > 0
• [project]_partP_tec_volume.dat if N < 0
– Sampling output file naming convention (one file per sampling
geometry G):
• [project]_tec_sampling_geomG_timestepT.dat if N > 0
• [project]_tec_sampling_geomG.dat if N < 0

FUN3D Training Workshop

http://fun3d.larc.nasa.gov 36
March 24-25, 2014
Boundary Output Visualization Example
&global
boundary_animation_freq = -1 Dump boundary vis at end of run
/
&boundary_output_variables
primitive_variables = .false. Turn off rho, u, v, w, p
cp = .true. Turn on Cp
yplus = .true. Turn on y+
/

FUN3D Training Workshop

http://fun3d.larc.nasa.gov 37
March 24-25, 2014
 Sampling Visualization Example
&sampling_parameters
number_of_geometries = 3 Want 3 sampling geometries
type_of_geometry(1) = 'plane‘ First geometry is a plane
plane_center(2,1) = -234.243 Plane y-coordinate
plane_normal(2,1) = 1.0 Plane y-normal
sampling_frequency(1) = -1 Write at end of run
type_of_geometry(2) = 'sphere’ Second geometry is a sphere
sphere_center(1,2) = 74.9 Center x-coordinate
sphere_center(2,2) = -107.7 Center y-coordinate
sphere_center(3,2) = 50.0 Center z-coordinate
sphere_radius(2) = 20.0 Sphere radius
sampling_frequency(2) = -1 Write at end of run
type_of_geometry(3) = 'isosurface’ Third geometry is an isosurface
isosurf_variable(3) = 'mach’ Isosurface will be based on Mach number
isosurf_value(3) = 1.00 Isosurface defined by Mach=1
sampling_frequency(3) = -1 Write at end of run
/
&sampling_output_variables
primitive_variables = .false. Turn off rho, u, v, w, p
mach = .true. Turn on Mach number
/

FUN3D Training Workshop

http://fun3d.larc.nasa.gov 38
March 24-25, 2014
 Volume Visualization Example
&global
volume_animation_freq = -1 Dump output at end of run
/
&volume_output_variables
export_to='tecplot’ Send results to Tecplot file
/

FUN3D Training Workshop

http://fun3d.larc.nasa.gov 39
March 24-25, 2014
 Slicing Visualization Example
&global
slice_freq = -1 Dump output at end of run
/
&slice_data
nslices = 1 Perform one slice
slice_location(1) = -234.243 Coordinate of slice
/

FUN3D Training Workshop

http://fun3d.larc.nasa.gov 40
March 24-25, 2014
 Troubleshooting/FAQ
• I can see what look like ragged dark lines on sampling surfaces and
volume data – what is that?
– Duplicate information at partition boundaries is not removed; if surface
is not completely opaque, double plotting locally doubles the
opaqueness (duplicate info is removed from boundary surface output)
– Turn off transparency in Tecplot
• When I dump out volume plot files in Tecplot format, I get a file for every
processor – is there a way around this?
– Not currently. However, Tecplot can be easily told to load all of the
files at once without having to individually select them all.
– The FUN3D team is working with Tecplot to develop their next
generation of I/O API’s, with special focus on massively parallel needs
– Alternative: switch to Fieldview or CGNS output, which uses a single
file

FUN3D Training Workshop

http://fun3d.larc.nasa.gov 41
March 24-25, 2014
 What We Learned
• Basic gridding requirements and file formats
• Runtime environment
• How to set up boundary conditions and very basic FUN3D input decks
• How to run a tetrahedral RANS solution for a wing-body VGRID mesh
• How to perform a 2D mixed element airfoil solution using an AFLR3
grid
• Some unhealthy things to watch for and possible remedies
• Overview of visualization output options and examples

Don’t hesitate to send questions our way!

[email protected]

FUN3D Training Workshop

http://fun3d.larc.nasa.gov 42
March 24-25, 2014